TopolearnThese three modules cover most photo-editing and data-augmentation needs without writing pixel code.
`ImageFilter` provides filters applied with im.filter(...):
GaussianBlur(radius),BoxBlur(radius)UnsharpMask(radius, percent, threshold): the standard sharpening tool;SHARPENis a simpler fixed kernelFIND_EDGES,CONTOUR,EMBOSS,SMOOTH,DETAILMedianFilter(size),MinFilter,MaxFilter: rank filters, good against speckle noiseKernel(size, kernel): your own 3x3 or 5x5 convolution
`ImageEnhance` classes each wrap an image and expose enhance(factor), where 1.0 returns the original, below 1 reduces and above 1 increases: Brightness, Contrast, Color (saturation; 0 gives grayscale) and Sharpness.
`ImageOps` has ready-made operations: grayscale, invert, autocontrast(im, cutoff=2) (stretch the histogram, ignoring the extreme 2%), equalize, posterize, solarize, expand(im, border, fill) for borders, pad, and colorize for tinting grayscale images.
All of these return new images. Some filters work only on certain modes: most need L or RGB, and palette images should be converted first. im.point(func_or_table) maps each pixel value through a function or lookup table, which is how you build custom curves efficiently.
Example
from PIL import Image, ImageEnhance, ImageFilter, ImageOps
im = Image.linear_gradient("L").resize((64, 64)).convert("RGB")
im.paste((200, 30, 30), (16, 16, 48, 48)) # red square
blur = im.filter(ImageFilter.GaussianBlur(3))
sharp = im.filter(ImageFilter.UnsharpMask(radius=2, percent=150, threshold=3))
edges = im.convert("L").filter(ImageFilter.FIND_EDGES)
print(blur.size == im.size, edges.mode) # True L
gray = ImageEnhance.Color(im).enhance(0.0) # saturation 0 = gray look
r, g, b = gray.getpixel((32, 32))
print(r == g == b) # True
darker = ImageEnhance.Brightness(im).enhance(0.5)
print(darker.getpixel((32, 32))) # (100, 15, 15)
dull = Image.new("L", (10, 1))
dull.putdata(list(range(100, 110)))
print(ImageOps.autocontrast(dull).getextrema()) # (0, 255)
print(ImageOps.invert(im).getpixel((32, 32))) # (55, 225, 225)
print(dull.point(lambda v: 255 if v > 104 else 0).getextrema()) # (0, 255)Remember
im.filter(ImageFilter.X)applies blurs, sharpening, edges, rank filtersImageEnhance.<Kind>(im).enhance(f): 1.0 = unchangedImageOpshas autocontrast, equalize, invert, expand, pad and moreim.pointmaps pixel values through a function or table
Watch out for
- Applying filters to a P-mode image and getting errors or odd results; convert to RGB or L first.